2021/10/20 by Pedro A. Ortega, Markus Kunesch, Ortega, Pedro A. +37 · 2 voices · 3 citations
Computer Science · #Artificial Intelligence (cs.AI) #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Reinforcement Learning in Robotics #Topic Modeling #cs.AI #cs.LG
paper · pdf · doi:10.48550/arxiv.2110.10819
openalex publication_date 2021/10/20 · arxiv published 2021/10/20 · arxiv updated 2021/10/20 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
The recent phenomenal success of language models has reinvigorated machine learning research, and large sequence models such as transformers are being applied to a variety of domains. One important problem class that has remained relatively elusive however is purposeful adaptive behavior. Currently there is a common perception that sequence models "lack the understanding of the cause and effect of their actions" leading them to draw incorrect inferences due to auto-suggestive delusions. In this report we explain where this mismatch originates, and show that it can be resolved by treating actions as causal interventions. Finally, we show that in supervised learning, one can teach a system to condition or intervene on data by training with factual and counterfactual error signals respectively.